File Access and Transfer

1
File Access and Transfer
2
Issues

• Access and transfer are different operations
• with different requirements
• Transfer
• move the file from one place to another
• binary or text
• recovery from interruption
• move whole file
• Access
• remove awareness of file location
• treat remote files as if they were local
• hide differences of file organization,
  characteristics
• manage issues of simultaneous access
• handle file integrity issues

3
File System Characteristics

• Structure
• refers to the logical and physical organization
  of data in the file
• Attributes
• are descriptive terms, provide information such
  as creation time, size, owner, etc.
• Operations
• are functions that may be applied to a file
  and/or its attributes
• create, delete, read, write, change access
  rights, etc.

4
A taxonomy of file system issues
Adapted from Distributed File Systems M.
Satyanarayanan 1989
Multiple users, multiple sites, multiple processes
Multiple users, one site, multiple processes
One user, one site, multiple processes
One user, one site, one process
Naming, Physical storage, Programming
interface, Integrity
Concurrency control, Serializability, Deadlocks
Security
Location
5
Taxonomy -1
One user, one site, one process
Naming, Physical storage, Programming
interface, Integrity

• Naming
• Is the name space flat or hierarchical? Tree
  structured? Cycles allowed? How long can the file
  names be? Are there file extensions or naming
  conventions that have semantic significance?
• Programming interface
• How do applications access the file system? Is
  there support for atomic actions?

6
Taxonomy - 2
One user, one site, one process
Naming, Physical storage, Programming
interface, Integrity

• Physical storage
• How is the file system abstraction mapped onto
  physical storage media? Is the programming
  interface independent of the storage media? How
  are the conflicting pulls between large blocks
  for performance benefits balanced with small
  blocks to reduce fragmentation?
• Integrity
• How are power, hardware, media, and software
  failures dealt with?

7
Taxonomy - 3
One user, one site, multiple processes
Concurrency control, Serializability, Deadlocks

• Concurrency control
• synchronization policies what granularity
  (entire file, bytes, blocks?) What locking
  modes? What combinations of locks can co-exist?
• Serializability
• An execution instance of interleaved transactions
  is serializable if there is some sequential
  execution order that yields the same results.
  How is serializability realized in this file
  system?
• Deadlocks
• How are deadlocks detected or avoided in this
  system?

8
Taxonomy -4
Multiple users, one site, multiple processes

• Security
• How are users identified and authenticated? Can
  groups share access to files? What privileges can
  be granted? How do those privileges map to file
  operations? Can the privileges be revoked?

Security
9
Taxonomy -5
Multiple users, multiple sites, multiple processes

• Location
• embedded in file names (as in full path)? Then
  moving a file also changes its name. Users may
  have problems finding their files. Location
  transparency hides the actual location of the
  file.
• Also involves availability and replication issues

Location
10
File Transfer

• Character or image
• ASCII is 7-bit code bit 8 used for parity check
• if bit 8 is part of the data, transfer will not
  work
• block size sender and receiver need to agree
• error detection sender and receiver must agree
• error correction most common is retransmission
• dialog control two-way or one-way by turns
• flow control match sender rate to receiver
  capability communication needed

11
FTP

• File Transfer Protocol
• workhorse of file movement
• whole file transfer, not general file access
• two versions 2-system, 3-system
• uses telnet to do login to the remote system
• anonymous ftp widely used for public distribution
  of files

12
Two-system FTP interaction
PI Protocol Interpreter DTP Data Transfer
Protocol
13
Three-system version of FTP
14
A modern interface to FTP
15
Model for generalized file access
16
General Hierarchical Structure
17
Flat file in the general hierarchical file
structure
18
Unstructured file in the general hierarchical
file structure
19
File Access, not just transfer

• FTAM (File Transfer, Access, and Management)
• Other versions created for specific applications
• FTAM defines regimes to capture various kinds of
  interaction
• association, file selection, file open, data
  transfer regimes
• enter and exit regimes according to operations
  needed

20
Summarizing

• Network-based application development requires
  access to files at remote locations
• file transfer brings the remote file to where it
  is needed. Operations are done on the local
  copy.
• General file access is more involved. Taxonomy
  of issues includes security, deadlock, integrity,
  interface differences, etc.
• General model of file interaction removes the
  details of individual file systems
• FTP is most common for moving files
• More is needed for general file access